Oncofusion

start / graph / hyp-occupancy-is-necessary-and-transcription-is-what-decides-whether-the-element-breaks

Hypothesismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q107-break-machinery-deposition

The element carries an equally strong CTCF/cohesin peak in HCT116 and in a B-lymphoblastoid line, yet breaks in one and not the other - so occupancy is necessary and not sufficient, and transcription of KMT2A is the switch

also called occupancy is necessary, transcription is sufficient · why HCT116 does not break · the reconciliation of Canela and Gothe at one element

The protein complex that marks this spot is just as strongly attached in bowel-cancer cells as in blood cells, but the DNA only snaps there in blood cells. The difference is probably whether the gene is switched on and being read.

The same thing, for a scientist

Two facts from this beat, each measured: **(1)** CTCF/RAD21 occupancy at *KMT2A* exon 11 is comparably strong in HCT116 (top 0.02% of chr11; summit GRCh38 chr11:118,488,694) and in GM12878 (113-135x genome mean, summit chr11:118,488,694 and 118,488,671) - **a 30 bp spread across three independent experiments in two cell types** [E1]. **(2)** Etoposide-induced DSBs at that element reach the 99.6th-99.95th percentile of chr11 in Nalm6, CD34+, K562, TK6 and MCF7, and only the 94.0th percentile in HCT116 - below the rank of HCT116's own untreated control [E1]. **Occupancy therefore does not predict breakage.** The hypothesis: **cohesin/CTCF occupancy is necessary but not sufficient, and the sufficient condition is active transcription through the element.** It makes Canela's *occupancy determines WHERE* and Gothe's *transcription determines HOW MUCH* into one statement instead of a contradiction - and it is the single-locus version of the reconciliation `mechanism-loop-anchor-dsb-hotspots` records as unresolved. **It already has one direct confirmation at this element**: DRB pre-treatment moves TK6 from the 98.016th to the 70.995th chr11 percentile there, with both replicates falling from 1.82-2.33x to 0.65x genome mean [E2]. **And its cheapest test is a lookup**: if *KMT2A* is lowly transcribed in lovastatin-arrested G1 HCT116, the HCT116 negative is explained and the hypothesis survives; if *KMT2A* is well transcribed there, the hypothesis fails and the explanation must be lineage- or cell-cycle-specific [E1].

Go deeper
The dissociation, stated so it can be attacked
cellCTCF/cohesin occupancy at the elementetoposide DSB rank at the element
HCT116 (colorectal)top 0.02% of chr11; CTCF 84.94x and RAD21 104.12x genome mean94.004th percentile - and 92.893rd untreated
GM12878 (B-lymphoblastoid)113-135x genome mean for CTCF, 35x for RAD21; summit at the same basenot measured (no break map in GM12878)
TK6 (lymphoblastoid)not measured directly; ENCODE-imported by Gothe99.610th percentile
Nalm6 (B-cell precursor)not measured99.948th percentile
CD34+ primary marrownot measured99.796th / 99.658th
K562not measured99.944th

The dissociation is only airtight for HCT116, which is the one cell with both measurements, and there it is stark: the strongest available occupancy and no break. For the other cells the occupancy is inferred from GM12878 or from ENCODE clusters. That is the honest state of the argument, and it is why the HCT116 transcription lookup is the decisive cheap test rather than a nice-to-have.

Why this is a reconciliation and not a third position

mechanism-loop-anchor-dsb-hotspots records the within-side contradiction: Canela 2017 says loop-anchor breakage is largely transcription-independent, Gothe 2019 says it scales with transcriptional output and requires elongation.

Both can be true of different quantities. Canela's claim is about the spatial distribution of breaks - where they fall, which is set by cohesin-dependent TOP2 loading, and which does not change much when you inhibit transcription because the anchors do not move. Gothe's claim is about the magnitude at a given anchor - how many breaks, which scales with elongation. A model in which occupancy sets the addressable set of sites and transcription sets the amplitude at each one produces exactly both observations, and it predicts the HCT116 result: the address is there, the amplitude is zero.

Canela 2019 itself is already halfway to this, separating TOP2 chromatin localisation and etoposide trapping (transcription-independent, requiring pre-existing cohesin) from conversion of a trapped complex into an irreversible DSB. This hypothesis says the conversion step is the transcription-dependent one, and that at KMT2A exon 11 the conversion step is what differs between HCT116 and a blood cell. Which is directly testable with the ExoT-versus-ExoVII END-seq pairing that Canela ran 18 times in mouse and never in human.

Predictions, in descending order of cost
  1. (one lookup) KMT2A nascent transcription or Pol II occupancy over exon 11 is low in lovastatin-arrested G1 HCT116 and high in TK6, K562, Nalm6 and CD34+ cells.
  2. (one read) In Gothe's deposited TK6 DRB arm, the element's rank falls further than the genome-wide shift. Partly confirmed already - 98.016th to 70.995th percentile, a 27-point rank drop - but no formal locus-versus-genome test was run.
  3. (one read) Across Gothe's three cell types, the element's DSB rank orders with KMT2A expression. Gothe reports the ~89-gene MLL recombinome is among the most highly expressed genes in these cells but does not rank the element against expression cell by cell.
  4. (one experiment) sBLISS in HCT116 +/- etoposide confirms the negative in a second assay, and sBLISS in HCT116 after forcing KMT2A transcription (CRISPRa at the promoter) creates the break where occupancy was already present. This is the hypothesis's own decisive experiment and it is cheaper than the degron.
  5. (the degron) In a blood cell, cohesin removal abolishes the break even while transcription continues - occupancy necessary. See exp-cohesin-degron-break-map-in-a-human-b-precursor-at-kmt2a.
Facts and figures
statementCTCF/cohesin occupancy at KMT2A canonical exon 11 is necessary but not sufficient for TOP2-induced breakage there; active transcription through the element is the switch that converts an occupied, TOP2-loaded anchor into a detectable double-strand break. Occupancy sets WHERE breaks can happen and transcription sets HOW MANY happen at each site - which makes Canela's transcription-independence and Gothe's transcription-dependence statements about two different quantities rather than a contradiction.
origindata-pattern
prior probability0.6
predictions
[
  "KMT2A nascent transcription or Pol II occupancy over exon 11 is low in lovastatin-arrested G1 HCT116 and high in TK6, K562, Nalm6 and CD34+ cells",
  "In Gothe's deposited TK6 DRB arm the element's rank falls further than the genome-wide shift - partly confirmed, 98.016th to 70.995th percentile, but no formal test run",
  "Across Gothe's three sBLISS cell types the element's DSB rank orders with KMT2A expression",
  "sBLISS in HCT116 reproduces the END-seq negative in a second assay; CRISPRa at the KMT2A promoter in HCT116 creates the break where occupancy is already present",
  "In a blood cell, acute cohesin removal abolishes the break while transcription continues - occupancy necessary",
  "An ExoT-only versus ExoVII+ExoT END-seq pairing in a human blood cell shows that occupancy governs the trapped-complex signal and transcription governs the conversion to an irreversible DSB"
]
experiments
[
  "exp-cohesin-degron-break-map-in-a-human-b-precursor-at-kmt2a"
]
analyses
[
  "ana-kmt2a-exon-11-element-is-a-top-0-05-percent-etoposide-dsb-site",
  "ana-hct116-is-the-wrong-cell-line-for-the-cohesin-arm-at-kmt2a",
  "absence-no-human-ctcf-or-cohesin-occupancy-measurement-in-a-blood-cell-at-kmt2a"
]
falsifiers
[
  "KMT2A is well transcribed in lovastatin-arrested G1 HCT116 - one ENCODE RNA-seq or Pol II lookup. Then occupancy and transcription are both present and the element still does not break, and the explanation must be lineage- or cell-cycle-specific rather than transcriptional.",
  "CTCF/RAD21 occupancy at this element turns out to be substantially WEAKER in HCT116 than in a blood cell once matched track types are compared - then occupancy does predict breakage after all and the hypothesis is unnecessary.",
  "DRB or triptolide in a blood cell fails to reduce breakage at this element beyond the genome-wide shift, tested with a proper locus-versus-genome null - then transcription is not the switch at this site even if it is genome-wide.",
  "A cohesin degron in a blood cell leaves the break intact while occupancy is lost - then occupancy is not necessary either, and the element's fragility is a property of something else entirely (sequence, nucleosome positioning, TOP2 recruitment independent of cohesin).",
  "The HCT116 negative fails to reproduce in a second assay - e.g. sBLISS in HCT116 shows a strong peak - in which case the whole dissociation is an END-seq protocol artefact of the 30-minute pulse."
]
How this could bite us back
  • If transcription is the switch, then every intervention that raises KMT2A expression - and menin inhibitors do change the KMT2A-fusion transcriptional programme - is in principle a fragility intervention at this element. That connects this beat to the menin-interface hypothesis in an uncomfortable direction and has not been measured.
  • The hypothesis also predicts that the highest-risk cell for a therapy-related KMT2A rearrangement is the one in which KMT2A is most highly transcribed, which in a paediatric context means the fetal and neonatal haematopoietic compartment - exactly where etoposide-era epidemiology found infant secondary leukaemia.
  • It is a reconciliation, and reconciliations are the most seductive kind of hypothesis in this graph because they make two papers agree. The dissociation it rests on has ONE cell type with both measurements and n = 1 in that cell type.
Still open
  1. Is KMT2A transcribed in lovastatin-arrested G1 HCT116? One ENCODE lookup decides whether this hypothesis survives its first contact with data, and this beat did not run it.
  2. Does the ExoT-only versus ExoVII+ExoT contrast behave as the hypothesis predicts - occupancy governing trapping, transcription governing conversion? Canela's own mouse pairs could be checked at the mouse Kmt2a orthologue without any new experiment, if the mouse element can be located by homology rather than by exon number.
  3. Triptolide, the orthogonal transcription inhibitor, has an END-seq arm in mouse activated B cells and has never been applied to the human element. Is the DRB result reproducible with a mechanistically different inhibitor?
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    CTCF/cohesin occupancy at KMT2A canonical exon 11 is comparably strong in HCT116 (top 0.02% of 918,411 chr11 147-bp bins for both CTCF and RAD21; 84.94x and 104.12x the respective file's genome mean; summits at GRCh38 chr11:118,488,694 and 118,488,677) and in the B-lymphoblastoid line GM12878 (CTCF 113.27x and 135.25x genome mean in two independent ENCODE experiments, RAD21 35.08x, maxima at GRCh38 chr11:118,488,694, 118,488,671 and 118,488,664), while etoposide-induced DSB signal at the same interval reaches the 99.948th percentile of chr11 in Nalm6, 99.944th in K562, 99.796th and 99.658th in two replicates of primary CD34+ marrow, 99.610th in TK6 and 99.883rd in MCF7, but only the 94.004th percentile in HCT116 - below the 92.893rd percentile occupied by HCT116's own untreated control once depth is accounted for.

    PMID 31202577DOI
    What this evidence cannot show

    The dissociation is airtight only for HCT116, the single cell type with both an occupancy and a break measurement at this element; for Nalm6, TK6, K562 and CD34+ the occupancy is inferred from GM12878 or from ENCODE cluster summaries, not measured. The HCT116 and GM12878 occupancy values are not directly comparable in magnitude - the HCT116 tracks are raw coverage and the ENCODE tracks are fold-change over control - so 'comparably strong' rests on the rank and summit agreement, not on the x-genome-mean numbers. HCT116 was lovastatin-arrested in G1 and pulsed for 30 min while every other cell was unsynchronised and treated for 4 h, so cell-cycle state and pulse length are confounded with cell type. All occupancy and break measurements are n = 1 except the two GM12878 CTCF experiments, the two CD34+ replicates and the two TK6 replicates.

    Evidence that argues against this (1)
    • The confound that most threatens this hypothesis is cell cycle, not transcription. Lovastatin-arrested G1 HCT116 has very little TOP2A - TOP2A is S/G2-enriched - and etoposide needs TOP2 to make a lesion at all. If the HCT116 negative is a TOP2A-abundance effect, the hypothesis's premise (equal occupancy, unequal breakage, transcription as the switch) is unfalsified but unsupported, because the dissociation would have a simpler explanation. Gothe's own data bear on this: MLL breakage in that paper depends on BOTH TOP2 isoforms in every cell tested, so a G1 cell retaining only TOP2B should still break, which weakens the confound - but it does not eliminate it. Gothe HJ et al. 2018
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Transcription dependence is directly demonstrated at this element in Gothe 2019's deposited sBLISS: DRB pre-treatment (200 uM, 3 h) before etoposide moves the KMT2A exon-11 interval from the 98.016th to the 70.995th percentile of 918,411 chr11 147-bp bins, with the depth-normalised enrichment falling from 2.33x and 1.82x (two untreated-then-etoposide replicates) to 0.65x and 0.65x (two DRB-then-etoposide replicates) and the maximum UMI-DSB pileup within the 1,048-bp element falling from 5 and 4 to 2 and 1; and Gothe's own Figure 5, titled 'The formation of MLL fusions is largely transcription-dependent', reports that DRB 'reduced the frequency of MLL fusions with the various partners to almost background levels' while not changing MLL-partner synapsis in undamaged cells.

    GEO GSE121740 samples GSM3444992, GSM3444993, GSM3444996 and GSM3444997 (TK6, DRB and untreated, with etoposide, two replicates) read over HTTP range requests with pybigtools 0.3.0 on 2026-09-22; and the Gothe full text, Results and Figure 5 legend · Gothe HJ, Bouwman BAM, Gusmao EG, Piccinno R, Sayols S, Drechsel O, Petrosino G, Minneker V, Josipovic N, Mizi A, Nielsen CF, Wagner EM, Takeda S, Sasanuma H, Hudson DF, Kindler T, Baranello L, Papantonis A, Crosetto N, Roukos V · NCBI GEO / bioRxiv (published as Mol Cell 75:267-283.e12) 2019
    PMID 31202576DOI
    What this evidence cannot show

    n = 2, and DRB reduces etoposide-induced DSBs globally in this dataset - Gothe's own aggregate plots show a genome-wide decrease across gene bodies and promoters - so part of the 27-percentile-point rank drop is the global effect. No formal locus-versus-genome-shift test was run, which is the single missing statistic. DRB inhibits CDK9 and therefore elongation everywhere, so this is not an element-specific perturbation. The fusion result is from the December 2018 preprint and is scored by imaging at whole-gene probe resolution.

    Evidence that argues against this (1)
    • DRB at 200 uM for 3 hours has effects beyond CDK9 - it inhibits casein kinase 2 and adenosine kinase, and prolonged elongation block changes chromatin accessibility and topological stress globally. So 'transcription-dependent' here means 'DRB-sensitive', and the cleaner complement, triptolide, appears exactly once in Gothe's text and was not used at MLL. Canela 2017 DID run a triptolide END-seq arm - GSM3714588-GSM3714591 equivalents in GSE99194, TRPexp_aB12h with and without triptolide - but only in mouse activated B cells, so the orthogonal transcription inhibitor and the human element have never met. NCBI Gene Expression Omnibus 2017
Connected to